Abstract: With an accumulation of abundant geo-information, production of high quality maps for presenting the geo-information challenges the capacity of GIS. Geo-database-driven cartography emerging as an effective means to produce maps from geo-databases, have not been scrutinized in a systematic way, especially in terms of modeling the process of map-making. A cartographic model being put forward in this paper characterizes the process of map-making by a transformation from geographical space to map space based on an algebra structure, and provides with a theoretical foundation and operational guideline for this merging technology. According to cartographic presentations of geo-data in topographic maps, geo-information is abstracted into two elements: geo-features and spatial relationships between the features, and further structured in form of algebra into two algebraic spaces; Geo-Feature Space in a geo-databases world and Map-Feature Space in a map world respectively. The spatial relationships in Geo-Feature Space are modeled by Euclidean space and the spatial relationships in Map-Feature Space are calibrated by the visual cognitive mechanism. The cartographic model is built up by transforming geo-features into map-features which is restrained by the consistent spatial relationships in the two spaces. By defining an equivalent-kernel transform among symbol cells, map symbolization based on feature classes is extended into that based on feature instances, fitting cartographic representations being both specified and universalized. With defining the displacement transform for geo-features, the process of map-making proceeds more smoothly, reducing being intervened to a large extent. This model through the two transforms, articulates the key issue for automating map-making and reveals the mechanism of the transformation of geo-information from geo-databases to map representation, providing a guideline for better design of functionality of GIS.